Exploring agents as evaluators instead of LLMs for AI testing

Original: what if instead of llm as judge, we use agent as judge, for evals?

Source: x.com ↗